Simultaneous Modeling of Cardiovascular Disease and Diabetes Incidence with Age, BMI, Smoking, and Physical Activity as Predictors

Abstract

This study investigated the simultaneous occurrence of cardiovascular disease (CVD) and diabetes and examined the linear and non-linear effects of age, body mass index (BMI), smoking status, and physical activity on both outcomes while accounting for their possible dependence. Data from 315 participants were analysed using descriptive statistics, correlation analysis, logistic regression, and Vector Generalized Additive Models (VGAMs). The prevalence of CVD and diabetes was 23.81% and 40.95%, respectively, and the co-occurrence of both conditions was 15.24%. Logistic regression showed that age and BMI were significant predictors of diabetes, whereas none of the selected predictors had a statistically significant linear association with CVD. The VGAM identified a significant non-linear relationship between age and diabetes incidence (p = 0.0134), while the non-linear effects of age and BMI on CVD were not statistically significant. The estimated residual correlation parameter was 0.552, indicating moderate positive dependence between CVD and diabetes after adjustment for covariates. The diabetes prediction model showed reasonable discriminatory ability (AUC = 0.76), whereas the CVD model showed poor discriminatory ability (AUC = 0.51). The VGAM joint model also produced lower AIC and BIC values than the restricted independence model, and the likelihood ratio test confirmed significant dependence between the two outcomes (LRT = 20.60, p < 0.001). These findings support the use of joint modelling for related cardiometabolic outcomes.